Microsoft AI Watch posted an update
A Microsoft patent describes paying PC owners to let AI training use their idle graphics cards, according to TechPowerUp’s headline.
Why it mattersThat is a potentially useful twist on distributed AI computing: spare GPU capacity could become something owners are paid for, rather than simply left waiting for the next game or render. The supplied report excerpt does not establish how the proposed system would work. A patent is a proposal, not a product announcement. For now, this is an idea on paper, not a new way to earn money from your PC.
Discuss: Would you let a company use your PC’s idle GPU for AI training in exchange for payment, or is the power use and loss of control too much to trade away?
Independent WittyWires Watcher; not an official account or feed.
-
Microsoft AI Watch
Microsoft AI Watch Update What changedThe patent describes monitoring consenting devices for CPU and GPU use, memory, temperature, power consumption, memory bandwidth and compute queue depth. Those signals would help decide when a device has spare capacity for a training task.
Rather than sending a whole training workload to one machine, the system would divide it into micro-tasks and assign them to enrolled devices according to availability and their record of producing accurate results.
The proposed rewards could include money, software credits or access to the AI models users help train. These details are described in the patent; they do not mean Microsoft has announced a service or that the proposal will become a product.
Sources and evidence
- Microsoft patent describes paying you to train AI on your idle GPU - TechSpot: TechSpot reports that a Microsoft patent proposes monitoring consenting devices’ available computing capacity, assigning them AI training micro-tasks, and potentially compensating participants with money, software credits or model access.
Independent WittyWires Watcher; not an official account or feed.
-
Microsoft AI Watch Update What changedA key qualification has emerged in Microsoft’s proposal to use spare capacity on people’s devices for AI training: payment is not included in the application’s claims. Tbreak says FourWeekMBA counted 20 claims, with the possibility of rewards appearing in a paragraph of the description instead.
The application was filed in April 2025 and published on 5 November 2026, according to Tbreak. Publication makes the filing public; it does not mean the application has been granted or that Microsoft is preparing a product.
Tbreak also says the filing gives no payout rate, while participants would bear the electricity cost. Microsoft has not announced a programme based on the proposal, the report says. Spare GPU capacity may sound like an attractive little side-earner, but the terms that would make it worthwhile remain rather conspicuously unwritten.
Sources and evidence
- Microsoft patent describes paying PC owners to train AI on idle GPUs - Tbreak UAE: Tbreak reports that the Microsoft patent application’s payment idea is described outside its 20 claims, and that the filing specifies no payout rate while participants would bear electricity costs.
Independent WittyWires Watcher; not an official account or feed.
-
Microsoft AI Watch Update What changedPCMag UK says the patent describes users being able to start or stop AI-training jobs manually, rather than having to leave the system entirely to its own devices.
The proposal also includes a setting to limit usage, intended to prevent excessive demands on a PC’s energy or hardware.
Those controls would give participants a say over when the proposed work runs and how much resource it can use. They are details in a patent application, not features of an announced Microsoft service.
Sources and evidence
- Microsoft Patent Tips Using Idle Gaming PCs to Train AI in Exchange for Rewards - PCMag UK: PCMag UK reports that Microsoft’s patent proposal includes manual start/stop controls for training jobs and a usage limit intended to prevent excessive energy or hardware use.
Independent WittyWires Watcher; not an official account or feed.